Extend #1869, update Intel SDE; Enable Linux AVX512 IFMA usage (#1871)

### Description of changes: This extends the work done on #1869. * Updates to latest version of Intel SDE; * Fixes minor bug in Go test script. * Disables use of AVX512 IFMA on Windows. By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and the ISC license. --------- Co-authored-by: Dan Pittman <[email protected]> Co-authored-by: Justin Smith <[email protected]>
aws · Sep 25, 2024 · 2df0b55 · 2df0b55
1 parent 266228a
commit 2df0b55
Show file tree

Hide file tree

Showing 9 changed files with 805 additions and 807 deletions.
diff --git a/.github/workflows/actions-ci.yml b/.github/workflows/actions-ci.yml
@@ -9,8 +9,8 @@ concurrency:
   cancel-in-progress: true
 env:
   GOPROXY: https://proxy.golang.org,direct
-  SDE_MIRROR_URL: "https://downloadmirror.intel.com/813591/sde-external-9.33.0-2024-01-07-win.tar.xz"
-  SDE_VERSION_TAG: sde-external-9.33.0-2024-01-07-win
+  SDE_MIRROR_URL: "https://downloadmirror.intel.com/831748/sde-external-9.44.0-2024-08-22-win.tar.xz"
+  SDE_VERSION_TAG: sde-external-9.44.0-2024-08-22-win
   PACKAGE_NAME: aws-lc
   # Used to enable ASAN test dimension.
   AWSLC_NO_ASM_FIPS: 1

diff --git a/crypto/fipsmodule/bn/asm/rsaz-2k-avx512.pl b/crypto/fipsmodule/bn/asm/rsaz-2k-avx512.pl
@@ -76,9 +76,7 @@
 
 if ($avx512ifma>0) {{{
 
-@_6_args_universal_ABI = $win64 ?
-("%rcx","%rdx","%r8","%r9","%r10","%r11") :
-("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
+@_6_args_universal_ABI = ("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
 
 ###############################################################################
 # void rsaz_amm52x20_x1_ifma256(BN_ULONG *res,

diff --git a/crypto/fipsmodule/bn/asm/rsaz-3k-avx512.pl b/crypto/fipsmodule/bn/asm/rsaz-3k-avx512.pl
@@ -74,9 +74,8 @@
 *STDOUT=*OUT;
 
 if ($avx512ifma>0) {{{
-@_6_args_universal_ABI = $win64 ?
-("%rcx","%rdx","%r8","%r9","%r10","%r11") :
-("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
+
+@_6_args_universal_ABI = ("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
 
 ###############################################################################
 # void rsaz_amm52x30_x1_ifma256(BN_ULONG *res,

diff --git a/crypto/fipsmodule/bn/asm/rsaz-4k-avx512.pl b/crypto/fipsmodule/bn/asm/rsaz-4k-avx512.pl
@@ -74,9 +74,7 @@
 *STDOUT=*OUT;
 
 if ($avx512ifma>0) {{{
-@_6_args_universal_ABI = $win64 ?
-("%rcx","%rdx","%r8","%r9","%r10","%r11") :
-("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
+@_6_args_universal_ABI = ("%rdi","%rsi","%rdx","%rcx","%r8","%r9");
 
 ###############################################################################
 # void rsaz_amm52x40_x1_ifma256(BN_ULONG *res,

diff --git a/crypto/fipsmodule/cpucap/internal.h b/crypto/fipsmodule/cpucap/internal.h
@@ -140,10 +140,12 @@ OPENSSL_INLINE int CRYPTO_is_VPCLMULQDQ_capable(void) {
 // 1100_0000_0010_0011_0000_0000_0000_0000
 #define CPU_CAP_AVX512IFMA_BITFLAGS 0xC0230000
 OPENSSL_INLINE int CRYPTO_is_AVX512IFMA_capable(void) {
-    return 0;
-// TODO: Re-enable once we understand Windows test failures.
-//  return (OPENSSL_ia32cap_get()[2] & CPU_CAP_AVX512IFMA_BITFLAGS) ==
-//         CPU_CAP_AVX512IFMA_BITFLAGS;
+#if defined(OPENSSL_WINDOWS)
+  return 0;
+#else
+  return (OPENSSL_ia32cap_get()[2] & CPU_CAP_AVX512IFMA_BITFLAGS) ==
+         CPU_CAP_AVX512IFMA_BITFLAGS;
+#endif
 }
 
 OPENSSL_INLINE int CRYPTO_is_VBMI2_capable(void) {